This paper presents the design of controller based on the principles of Neural networks. The concept of artificial intelligent techniques greatly helps in overcoming the disadvantages posed by the conventional controllers. A hierarchical architecture of three layer feed forward neural network (NN) is proposed for controller design based on back propagation algorithm (BPA). Area Control Error (ACE) is considered as input to the neural network controller and the output of the controller is provided to the governor in each area. The proposed controllers are tested for a two area hydrothermal system. Simulation results show that the limitations of conventional controller can be overcome by including Neural concept and thereby the dynamic response of the system with respect to peak time, overshoot and settling time can be improved drastically.